Stock-Price-Prediction

This repository contains code and resources for predicting stock prices using machine learning techniques. The goal is to develop a model that can accurately predict future stock prices based on historical data.

Table of Contents

• Introduction
• Installation
• Data
• Usage
• Results
• Contributing
• License
• Contact

Introduction

Stock price prediction is a challenging task due to the volatility and complexity of financial markets. This project aims to leverage machine learning algorithms to make accurate predictions based on historical stock price data.

Installation

To run this project locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/Shengwei-Peng/Stock-Price-Prediction.git
   cd Stock-Price-Prediction

2. Install the required dependencies:

   pip install -r requirements.txt

Data

The dataset used in this project consists of historical stock prices fetched using the twstock library. Here is an example of how to fetch the data:

import twstock
import pandas as pd

# Fetch historical stock data for stock code '2330' from January 2014 to present
stock = twstock.Stock('2330')
data = pd.DataFrame(stock.fetch_from(2014, 1))

# Save the data to a CSV file
data.to_csv("2330.csv", index=False)

Below is a brief description of each column in the dataset and a sample table to illustrate the data format:

Column	Description	Data Type	Unit
Date	The trading date	YYYY-MM-DD
Capacity	The total traded volume of stocks	int	Shares
Turnover	The total value of traded stocks	int	TWD
Open	The opening price of the stock	float	TWD
High	The highest price during the day	float	TWD
Low	The lowest price during the day	float	TWD
Close	The closing price of the stock	float	TWD
Change	The price change compared to the previous day	float	TWD
Transaction	The number of transactions	int	Transactions

Example Data

Date	Capacity	Turnover	Open	High	Low	Close	Change	Transaction
2014-01-02	15133724	1578860796	105.0	105.5	103.5	104.5	-1.0	4136
2014-01-03	41160529	4220062918	103.0	103.0	102.0	102.5	-2.0	8524
2014-01-06	23729719	2432745661	102.0	103.0	102.0	102.5	0.0	4874

Usage

To use the stock price prediction tool, follow these steps:

1. Prepare your dataset: Ensure your dataset is in the correct format and located in the data directory.
2. Run the main script: Execute the following command to start the process:

   python main.py --data_path data/2330.csv --model transformer --test_size 20 --window_size 20 --seed 0

The script will perform the following steps:

• Preprocessing: Clean and prepare the data for analysis.
• Training: Train the prediction models on the prepared data.
• Evaluation: Assess the performance of the trained models and generate performance metrics.
• Visualization: Create visualizations to compare the predicted stock prices with actual prices.

Configuration Options

You can customize the behavior of the tool using various configuration options. Here are some examples:

• data_path: Specify the path to the input data file.
• model: Choose the machine learning model to use. Options include linear_regression, svm, random_forest, xgboost, lstm and transformer.
• test_size: Specify the number of latest trading days to use as the test set.
• window_size: Specify the number of past days (D) to use for predicting the next day (D+1).
• seed: Set a random seed for reproducibility of results.

For a complete list of arguments, refer to the parse_args function in utils.py.

Visualizing Results

The tool generates various plots to help understand the performance of the models. It includes:

• Line plots: Comparing the predicted and actual stock prices over time.
• Residual plots: Visualizing the residuals (differences between true and predicted values) over time to identify any patterns or anomalies.
• Scatter plots: Comparing the true prices with the predicted prices to assess the model's accuracy visually.

These visualizations help in analyzing the accuracy and effectiveness of the predictions.

Results

The performance of each model is evaluated using various metrics, including R², MSE, RMSE, MAE and NDEI. Below are the evaluation results for the implemented models:

Model	R²	MSE	RMSE	MAE	NDEI
Linear Regression	0.5790	25.5890	5.0586	4.3584	0.6488
SVM	0.4966	30.5983	5.5316	4.2504	0.7095
Random Forest	0.7832	13.1793	3.6303	3.0765	0.4656
XGBoost	0.6167	23.2974	4.8267	3.8451	0.6191
LSTM	0.7675	14.1318	3.7592	3.0583	0.4822
Transformer	0.7645	14.3175	3.7838	3.0347	0.4853

Evaluation Metrics

• R² (Coefficient of Determination): Indicates how well the predictions match the actual values. A negative R² indicates that the model is performing worse than a horizontal line (mean prediction).
• MSE (Mean Squared Error): Measures the average squared difference between the predicted and actual values. Lower values indicate better performance.
• RMSE (Root Mean Squared Error): The square root of MSE, providing an error metric in the same unit as the target variable. Lower values indicate better performance.
• MAE (Mean Absolute Error): Measures the average absolute difference between the predicted and actual values. Lower values indicate better performance.
• NDEI (Non-dimensional Error Index): The ratio of the RMSE to the standard deviation of the actual values. Lower values indicate better performance, and it is useful for comparing the error relative to the variability of the data.

Data Details

• Stock: TSMC (2330)
• Date Range: January 2014 to December 2023
• Test Set: The last 20 trading days

The models were trained on data spanning nearly a decade to capture various market conditions and trends. The final 20 trading days of this period were reserved for testing to evaluate the models' performance on recent data.

Contributing

Contributions are welcome! Please follow these steps to contribute:

1. Fork the repository.
2. Create a new branch (git checkout -b feature-branch).
3. Commit your changes (git commit -am 'Add new feature').
4. Push to the branch (git push origin feature-branch).
5. Create a new Pull Request.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contact

For any questions or inquiries, please contact [email protected]